Cell image analysis method and system based on geometric mean and graph structure

Abstract

The invention discloses a cell image analysis method and system based on a geometric mean and a graph structure, and the method comprises the steps: inputting a multi-channel fluorescence image into a cell recognition algorithm model, and obtaining a cell distribution graph; calculating a cell deviation degree parameter in the cell distribution diagram according to a geometric mean algorithm; based on a graph neural network, extracting cell graph features corresponding to the cell distribution graph and the cell deviation degree parameter; and according to the time sequence corresponding to the cell map features at the plurality of historical time points, predicting cell pathology classification and change trend at a future time point based on an LSTM neural network. Therefore, the method can achieve the accurate prediction of the cell pathology classification and the evolution trend of the cell pathology classification at the future time point, improves the automation level and the prospective judgment capability of cell pathology analysis, and provides efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a cell image analysis method and system based on geometric mean and graph structure. Background Art

[0002] In the prior art, cell pathology analysis mainly relies on manual observation of fluorescent images or the use of traditional image processing algorithms to segment and count cell images. However, traditional methods have poor robustness and limited feature extraction capabilities when processing multi-channel fluorescent images, making it difficult to accurately identify the complex distribution relationships between cells. In addition, existing analysis methods generally lack the ability to model and predict the changing trends of cell distribution, and are unable to combine historical image data for time-series evolution analysis, resulting in a lack of forward-looking judgment on the development of cell pathology, making it difficult to meet the actual needs of disease early warning and dynamic tracking. It can be seen that the prior art has defects that need to be addressed urgently. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a cell image analysis method and system based on geometric mean and graph structure, which can realize accurate prediction of cell pathology classification and its evolution trend at future time points, improve the automation level and forward-looking judgment ability of cell pathology analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

[0004] In order to solve the above technical problems, the first aspect of the present invention discloses a cell image analysis method based on geometric mean and graph structure, the method comprising: Input the multi-channel fluorescence image into the cell recognition algorithm model to obtain the cell distribution map; Calculating the cell deviation parameter in the cell distribution map according to the geometric mean algorithm; Based on a graph neural network, extracting cell map features corresponding to the cell distribution map and the cell deviation parameter; According to the time series corresponding to the cell map features at multiple historical time points, based on the LSTM neural network, the cell pathology classification and change trend at future time points are predicted.

[0005] As an optional embodiment, in the first aspect of the present invention, the step of inputting the multi-channel fluorescence image into a cell recognition algorithm model to obtain a cell distribution map comprises: The cell fluorescence image of each channel is input into the trained cell image segmentation algorithm model to obtain the cell distribution map in the cell fluorescence image of each channel.

[0006] As an optional embodiment, in the first aspect of the present invention, The cell distribution map includes a plurality of cell images and a cell type corresponding to each cell image.

[0007] As an optional embodiment, in the first aspect of the present invention, the cell deviation parameter in the cell distribution map is calculated according to the geometric mean algorithm, comprising: For each cell image in the cell distribution map, calculating the geometric mean of the image deviation parameters of the cell image in the cell fluorescence images of all channels to obtain the image average parameter corresponding to the cell image; Calculating the geometric mean of the position distances between the image position of the cell image and the positions of multiple reference cells to obtain a position average parameter corresponding to the cell image; The weighted average value of the image average parameter and the position average parameter is calculated to obtain the cell deviation parameter corresponding to the cell image.

[0008] As an optional embodiment, in the first aspect of the present invention, the image deviation parameter is the difference between the image parameter of the cell image and the standard parameter value; the image parameter includes at least one of image brightness, image color value, image morphology, image texture and image spatial distribution parameter.

[0009] As an optional embodiment, in the first aspect of the present invention, the reference cell position is the position of the nearest cell image of a preset dangerous cell type, the position of the nearest cell image of the same cell type, or the position of a cell image at the center of a preset cell area.

[0010] As an optional embodiment, in the first aspect of the present invention, the extracting the cell distribution map and the cell map features corresponding to the cell deviation parameter based on the graph neural network includes: Each cell image in the cell distribution map is used as a map node, and the cell deviation parameter corresponding to the cell image corresponding to each map node is used as a map node characteristic parameter; For any two of the graph nodes, calculating the ratio between the cell deviation parameters respectively corresponding to the two graph nodes, and obtaining the node adjacency feature between the two graph nodes; Determine all the graph nodes and the node adjacency features corresponding to each other as a graph network corresponding to the cell distribution graph; The graph network is input into a trained graph neural network to obtain cell graph features corresponding to the graph network; the graph neural network is trained by a training data set including multiple training cell graph networks and corresponding feature annotations.

[0011] As an optional embodiment, in the first aspect of the present invention, the method of predicting the cell pathology classification and change trend at a future time point based on the time series corresponding to the cell map features at multiple historical time points and based on the LSTM neural network includes: The cell map features of the same patient at multiple historical time points are sorted from early to late based on the time points to obtain a feature time series sequence; Inputting the characteristic time series into the trained LSTM neural network to obtain multiple cell pathology types corresponding to the output future time points; According to the type changes of the multiple cell pathology types corresponding to the future time point, the pathology change trend corresponding to the patient is determined.

[0012] A second aspect of an embodiment of the present invention discloses a cell image analysis system based on geometric mean and graph structure, the system comprising: The recognition module is used to input the multi-channel fluorescence image into the cell recognition algorithm model to obtain the cell distribution map; A calculation module, used for calculating the cell deviation parameter in the cell distribution map according to a geometric mean algorithm; An extraction module, used for extracting the cell distribution map and the cell map features corresponding to the cell deviation parameter based on a graph neural network; The prediction module is used to predict the cell pathology classification and change trend at future time points based on the time series corresponding to the cell map features at multiple historical time points and based on the LSTM neural network.

[0013] As an optional embodiment, in the second aspect of the present invention, the recognition module inputs the multi-channel fluorescence image into the cell recognition algorithm model to obtain the specific method of the cell distribution map, including: The cell fluorescence image of each channel is input into the trained cell image segmentation algorithm model to obtain the cell distribution map in the cell fluorescence image of each channel.

[0014] As an optional embodiment, in the second aspect of the present invention, The cell distribution map includes a plurality of cell images and a cell type corresponding to each cell image.

[0015] As an optional embodiment, in the second aspect of the present invention, the specific manner in which the calculation module calculates the cell deviation parameter in the cell distribution map according to the geometric mean algorithm includes: For each cell image in the cell distribution map, calculating the geometric mean of the image deviation parameters of the cell image in the cell fluorescence images of all channels to obtain the image average parameter corresponding to the cell image; Calculate the geometric mean of the positional distances between the image position of the cell image and multiple reference cell positions to obtain the positional average parameter corresponding to the cell image; Calculate the weighted summation average of the image average parameter and the positional average parameter to obtain the cell deviation parameter corresponding to the cell image.

[0016] As an optional implementation manner, in the second aspect of the present invention, the image deviation parameter is the difference between the image parameter of the cell image and the standard parameter value; the image parameter includes at least one of image brightness, image color value, image morphology, image texture, and image spatial distribution parameter.

[0017] As an optional implementation manner, in the second aspect of the present invention, the reference cell position is the position of the cell image of the nearest preset dangerous cell type, the position of the cell image of the nearest same cell type, or the position of the cell image at the center of the preset cell area.

[0018] As an optional implementation manner, in the second aspect of the present invention, the specific manner in which the extraction module extracts the cell graph features corresponding to the cell distribution map and the cell deviation parameter based on a graph neural network includes: Use each cell image in the cell distribution map as a graph node, and use the cell deviation parameter corresponding to the cell image corresponding to each graph node as the graph node feature parameter; For any two of the graph nodes, calculate the ratio between the cell deviation parameters respectively corresponding to the two graph nodes to obtain the node adjacency feature between the two graph nodes; Determine all the graph nodes and the corresponding node adjacency features therebetween as the graph network corresponding to the cell distribution map; Input the graph network into a trained graph neural network to obtain the cell graph features corresponding to the graph network; the graph neural network is trained through a training data set including multiple training cell graph networks and corresponding feature annotations.

[0019] As an optional implementation manner, in the second aspect of the present invention, the specific manner in which the prediction module predicts the cell pathological classification and change trend at a future time point based on the LSTM neural network according to the time series corresponding to the cell graph features at multiple historical time points includes: Sort the cell graph features of the same patient at multiple historical time points from earliest to latest to obtain a feature time series; Input the feature time series into a trained LSTM neural network to obtain multiple cell pathological types corresponding to the output future time point; Determine the pathological change trend corresponding to the patient according to the type changes of multiple cytopathological types corresponding to the future time point.

[0020] The third aspect of the present invention discloses another cell image analysis system based on geometric mean and graph structure. The system includes: A memory storing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory and executes some or all of the steps in the cell image analysis method based on geometric mean and graph structure disclosed in the first aspect of the present invention.

[0021] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which are used to execute some or all of the steps in the cell image analysis method based on geometric mean and graph structure disclosed in the first aspect of the present invention when called.

[0022] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: In the present invention, a multi-channel fluorescence image is input into a cell recognition algorithm model to obtain a cell distribution map, and a geometric mean algorithm is used to calculate the cell deviation parameter in the map. On this basis, a graph neural network is combined to extract graph structure features integrating the cell distribution map and the deviation parameter, and a time series sequence is constructed and input into an LSTM neural network for cell pathological classification and prediction of change trends, so as to accurately predict the cell pathological classification and its evolution trend at a future time point, improve the automation level and forward-looking judgment ability of cell pathological analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0024] Figure 1 It is a flowchart of a cell image analysis method based on geometric mean and graph structure disclosed in an embodiment of the present invention.

[0025] Figure 2 It is a structural diagram of a cell image analysis system based on geometric mean and graph structure disclosed in an embodiment of the present invention.

[0026] Figure 3It is a schematic structural diagram of another cell image analysis system based on geometric mean and graph structure disclosed in the embodiments of the present invention. Detailed implementation manners

[0027] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0028] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0029] Referring to "embodiment" herein means that a specific feature, structure or characteristic described in connection with the embodiment can be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0030] The present invention discloses a cell image analysis method and system based on geometric mean and graph structure. By inputting a multi-channel fluorescence image into a cell recognition algorithm model to obtain a cell distribution map, and using the geometric mean algorithm to calculate the cell deviation parameter in the map. On this basis, a graph neural network is combined to extract graph structure features that fuse the cell distribution map and the deviation parameter, and a time series sequence is constructed and input into an LSTM neural network for cell pathology classification and prediction of change trends, so as to accurately predict the cell pathology classification and its evolution trend at future time points, improve the automation level and forward-looking judgment ability of cell pathology analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis. The following will be described in detail respectively.

[0031] Embodiment 1 Please refer to Figure 1 , Figure 1 It is a schematic flowchart of a cell image analysis method based on geometric mean and graph structure disclosed in the embodiments of the present invention. Among them,Figure 1 The described cell image analysis method based on geometric mean and graph structure can be applied to a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). As Figure 1 shown, the cell image analysis method based on geometric mean and graph structure may include the following operations: 101. Input a multi-channel fluorescence image into a cell recognition algorithm model to obtain a cell distribution map.

[0032] 102. Calculate the cell deviation parameter in the cell distribution map according to the geometric mean algorithm. 103. Based on a graph neural network, extract the cell graph features corresponding to the cell distribution map and the cell deviation parameter. 104. According to the time series corresponding to the cell graph features at multiple historical time points, based on the LSTM neural network, predict the cell pathological classification and change trend at a future time point.

[0033] It can be seen that in the above-mentioned invention embodiments, by inputting a multi-channel fluorescence image into a cell recognition algorithm model to obtain a cell distribution map, and using the geometric mean algorithm to calculate the cell deviation parameter in the map, on this basis, combining a graph neural network to extract the graph structure features that fuse the cell distribution map and the deviation parameter, and by constructing a time series and inputting it into the LSTM neural network for cell pathological classification and change trend prediction, it is possible to accurately predict the cell pathological classification and its evolution trend at a future time point, improve the automation level and forward-looking judgment ability of cell pathological analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

[0034] As an optional embodiment, in the above steps, inputting a multi-channel fluorescence image into a cell recognition algorithm model to obtain a cell distribution map includes: Input the cell fluorescence image of each channel into a trained cell image segmentation algorithm model to obtain the cell distribution map in the cell fluorescence image of each channel.

[0035] It can be seen that through the above optional embodiment, by inputting the cell fluorescence image of each channel into a trained cell image segmentation algorithm model, it is possible to accurately extract the cell regions and distribution information corresponding to each channel from the original fluorescence image, which not only improves the accuracy and robustness of cell image segmentation, but also avoids the recognition errors caused by image overlap and signal interference in multi-channel image processing by traditional methods, thereby providing a clearly structured and channel-independent image basis for subsequent cell morphology analysis and state recognition, effectively improving the overall image processing efficiency and analysis accuracy.

[0036] As an optional embodiment, in the above steps, The cell distribution map includes multiple cell images and the cell types corresponding to each cell image.

[0037] It can be seen that through the above optional embodiments, the content of the cell distribution map is defined to comprehensively characterize the cell distribution characteristics, assist in accurately predicting the cell pathological classification and its evolution trend at future time points, improve the automation level and forward-looking judgment ability of cell pathological analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

[0038] As an optional embodiment, in the above steps, according to the geometric mean algorithm, calculating the cell deviation parameter in the cell distribution map includes: For each cell image in the cell distribution map, calculate the geometric mean of the image deviation parameters of the cell image in the cell fluorescence images of all channels to obtain the image average parameter corresponding to the cell image; Calculate the geometric mean of the position distances between the image position of the cell image and multiple reference cell positions to obtain the position average parameter corresponding to the cell image; Calculate the weighted sum average of the image average parameter and the position average parameter to obtain the cell deviation parameter corresponding to the cell image.

[0039] It can be seen that through the above optional embodiments, for each cell image in the cell distribution map, calculating the geometric mean of the image deviation parameters of the cell image in the cell fluorescence images of all channels can uniformly measure the overall deviation of the same cell in terms of morphology or signal expression in multi-channel images; further, by calculating the geometric mean of the position distances between the position of the cell image and multiple reference cell positions, the degree of outlier in its spatial distribution is evaluated; finally, the image average parameter and the position average parameter are weighted and summed and averaged to obtain a more comprehensively representative cell deviation parameter, realizing the fusion quantification of the deviation degree of a single cell in two dimensions of image signal and spatial position, and providing key index support for abnormal cell recognition and subsequent pathological trend modeling.

[0040] As an optional embodiment, in the above steps, the image deviation parameter is the difference between the image parameter of the cell image and the standard parameter value; the image parameter includes at least one of image brightness, image color value, image morphology, image texture, and image spatial distribution parameter.

[0041] It can be seen that through the above optional embodiments, the details of the image deviation parameter are defined to facilitate the subsequent accurate characterization of the image deviation parameter of the cell, assist in accurately predicting the cell pathological classification and its evolution trend at future time points, improve the automation level and forward-looking judgment ability of cell pathological analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

[0042] As an alternative embodiment, in the above steps, the reference cell position is the position of the cell image of the nearest preset dangerous cell type, the position of the cell image of the same cell type that is the nearest, or the position of the cell image at the center of the preset cell region.

[0043] It can be seen that through the above alternative embodiments, the types of reference cell positions are defined, which facilitates the subsequent accurate calculation of the position average parameters of cells, assists in accurately predicting the cell pathological classification and its evolution trend at future time points, improves the automation level and forward-looking judgment ability of cell pathological analysis, and provides efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

[0044] As an alternative embodiment, in the above steps, based on a graph neural network, cell map features corresponding to the cell distribution map and the cell deviation parameter are extracted, including: Taking each cell image in the cell distribution map as a graph node, and taking the cell deviation parameter corresponding to the cell image corresponding to each graph node as the graph node feature parameter; For any two graph nodes, calculate the ratio between the cell deviation parameters corresponding to the two graph nodes respectively to obtain the node adjacency feature between the two graph nodes; Determine all graph nodes and the corresponding node adjacency features between them as the graph network corresponding to the cell distribution map; Input the graph network into the trained graph neural network to obtain the cell map features corresponding to the graph network; the graph neural network is trained through a training data set including multiple training cell graph networks and corresponding feature annotations.

[0045] It can be seen that through the above alternative embodiments, taking each cell image in the cell distribution map as a graph node and using the cell deviation parameter corresponding to each graph node as the graph node feature parameter can fuse the morphological and position information of single cells in the cell graph structure; further, by calculating the ratio between the cell deviation parameters of any two graph nodes and extracting the node adjacency features between them, it helps to describe the relative differences in the deviation degree between cells; and then constructing a cell graph network and inputting it into the trained graph neural network can effectively extract the deep features of the cell population distribution pattern, realize the comprehensive characterization of cell states in the spatial structure dimension, and improve the sensitivity and recognition ability for local abnormalities and pathological trends.

[0046] As an alternative embodiment, in the above steps, based on the time series corresponding to the cell map features at multiple historical time points, the cell pathological classification and change trend at future time points are predicted based on an LSTM neural network, including: Sort the cell map features of the same patient at multiple historical time points from the earliest to the latest to obtain a feature time series; Input the characteristic time series into the trained LSTM neural network to obtain multiple cell pathological types corresponding to the future time points of the output; Determine the pathological change trend corresponding to the patient according to the type changes of multiple cell pathological types corresponding to the future time points.

[0047] It can be seen that through the above optional embodiments, through the continuous modeling of time series features and the trend prediction based on deep neural networks, the accurate recognition ability of pathological evolution trends and the dynamic adaptability of the model to the pathological development process are improved, and the foresight and reliability of cell pathological change analysis are enhanced.

[0048] Embodiment 2 Please refer to Figure 2 , Figure 2 which is a schematic structural diagram of a cell image analysis system based on geometric mean and graph structure disclosed in an embodiment of the present invention. Among them, Figure 2 The described cell image analysis system based on geometric mean and graph structure can be applied to a data processing system / data processing device / data processing server (wherein, the server includes a local processing server or a cloud processing server). As Figure 2 shown, the cell image analysis system based on geometric mean and graph structure may include: An identification module 201, configured to input a multi-channel fluorescence image into a cell recognition algorithm model to obtain a cell distribution map.

[0049] A calculation module 202, configured to calculate a cell deviation parameter in the cell distribution map according to the geometric mean algorithm. An extraction module 203, configured to extract cell graph features corresponding to the cell distribution map and the cell deviation parameter based on a graph neural network. A prediction module 204, configured to predict the cell pathological classification and change trend at a future time point based on the time series corresponding to the cell graph features at multiple historical time points and based on an LSTM neural network.

[0050] It can be seen that in the above embodiments of the invention, a cell distribution map is obtained by inputting a multi-channel fluorescence image into a cell recognition algorithm model, and the cell deviation parameter in the map is calculated by using the geometric mean algorithm. On this basis, a graph neural network is combined to extract graph structure features that fuse the cell distribution map and the deviation parameter, and a time series is constructed and input into an LSTM neural network for cell pathological classification and change trend prediction, so as to be able to accurately predict the cell pathological classification and its evolution trend at a future time point, improve the automation level and foresight judgment ability of cell pathological analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

[0051] As an alternative embodiment, the specific manner in which the recognition module inputs the multi-channel fluorescence image into the cell recognition algorithm model to obtain the cell distribution map includes: Input the cell fluorescence image of each channel into the trained cell image segmentation algorithm model to obtain the cell distribution map in the cell fluorescence image of each channel.

[0052] It can be seen that through the above alternative embodiment, by inputting the cell fluorescence image of each channel into the trained cell image segmentation algorithm model, the cell regions and distribution information corresponding to each channel can be accurately extracted from the original fluorescence image, which not only improves the accuracy and robustness of cell image segmentation, but also avoids the recognition errors caused by image overlap and signal interference in multi-channel image processing of traditional methods. Thus, it provides a clearly structured and channel-independent image basis for subsequent cell morphology analysis and status recognition, effectively improving the overall image processing efficiency and analysis accuracy.

[0053] As an alternative embodiment, The cell distribution map includes multiple cell images and the cell type corresponding to each cell image.

[0054] It can be seen that through the above alternative embodiment, the content of the cell distribution map is defined to comprehensively characterize the cell distribution characteristics, assist in accurately predicting the cell pathological classification and its evolution trend at future time points, improve the automation level and prospective judgment ability of cell pathological analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

[0055] As an alternative embodiment, the specific manner in which the calculation module calculates the cell deviation parameter in the cell distribution map according to the geometric mean algorithm includes: For each cell image in the cell distribution map, calculate the geometric mean of the image deviation parameters of the cell image in the cell fluorescence images of all channels to obtain the image average parameter corresponding to the cell image; Calculate the geometric mean of the position distances between the image position of the cell image and multiple reference cell positions to obtain the position average parameter corresponding to the cell image; Calculate the weighted sum average of the image average parameter and the position average parameter to obtain the cell deviation parameter corresponding to the cell image.

[0056] It can be seen that through the above optional embodiments, for each cell image in the cell distribution map, calculating the geometric mean of the image deviation parameters of the cell image in the cell fluorescence images of all channels can uniformly measure the overall deviation of the same cell in terms of morphology or signal expression in multi-channel images; further, by calculating the geometric mean of the position distances between the position of the cell image and the positions of multiple reference cells, the degree of outlier in its spatial distribution is evaluated; finally, the image average parameter and the position average parameter are weighted and summed and averaged to obtain a more comprehensively representative cell deviation parameter, realizing the fusion quantification of the deviation degree of a single cell in two dimensions of image signal and spatial position, and providing key index support for abnormal cell recognition and subsequent pathological trend modeling.

[0057] As an optional embodiment, the image deviation parameter is the difference between the image parameter of the cell image and the standard parameter value; the image parameter includes at least one of image brightness, image color value, image morphology, image texture, and image spatial distribution parameter.

[0058] It can be seen that through the above optional embodiments, the details of the image deviation parameter are defined, so as to accurately characterize the image deviation parameter of the cell subsequently, assist in realizing the accurate prediction of cell pathological classification and its evolution trend at future time points, improve the automation level and forward-looking judgment ability of cell pathological analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

[0059] As an optional embodiment, the reference cell position is the position of the cell image of the nearest preset dangerous cell type, the position of the cell image of the nearest same cell type, or the position of the cell image at the center of the preset cell region.

[0060] It can be seen that through the above optional embodiments, the types of reference cell positions are defined, so as to accurately calculate the position average parameter of the cell subsequently, assist in realizing the accurate prediction of cell pathological classification and its evolution trend at future time points, improve the automation level and forward-looking judgment ability of cell pathological analysis, and provide efficient and reliable technical support for disease monitoring and auxiliary diagnosis.

[0061] As an optional embodiment, the specific manner in which the extraction module extracts the cell map features corresponding to the cell distribution map and the cell deviation parameter based on the graph neural network includes: Taking each cell image in the cell distribution map as a graph node, and taking the cell deviation parameter corresponding to the cell image corresponding to each graph node as the graph node feature parameter; For any two graph nodes, calculating the ratio between the cell deviation parameters respectively corresponding to the two graph nodes to obtain the node adjacency feature between the two graph nodes; All graph nodes and the corresponding node adjacency features between them are determined as the graph network corresponding to the cell distribution map; The graph network is input into the trained graph neural network to obtain the cell graph features corresponding to the graph network; the graph neural network is trained through a training data set including multiple training cell graph networks and corresponding feature annotations.

[0062] It can be seen that through the above optional embodiments, taking each cell image in the cell distribution map as a graph node and using the cell deviation parameter corresponding to each graph node as the graph node feature parameter, the morphology and position information of single cells can be fused in the cell graph structure; further, by calculating the ratio between the cell deviation parameters of any two graph nodes, the node adjacency features between them are extracted, which helps to depict the relative differences in the deviation degree between cells; furthermore, by constructing a cell graph network and inputting it into the trained graph neural network, the deep features of the cell population distribution pattern can be effectively extracted, realizing the comprehensive characterization of cell states in the spatial structure dimension, and enhancing the sensitivity and recognition ability for local abnormalities and pathological trends.

[0063] As an optional embodiment, the specific manner in which the prediction module predicts the cell pathological classification and change trend at a future time point based on the time series corresponding to the cell graph features at multiple historical time points and based on the LSTM neural network includes: Sort the cell graph features of the same patient at multiple historical time points from the earliest to the latest to obtain a feature time series; Input the feature time series into the trained LSTM neural network to obtain multiple cell pathological types corresponding to the output future time point; Determine the pathological change trend corresponding to the patient according to the type changes of multiple cell pathological types corresponding to the future time point.

[0064] It can be seen that through the above optional embodiments, through the continuous modeling of time series features and trend prediction based on deep neural networks, the accurate recognition ability of pathological evolution trends and the dynamic adaptability of the model to the pathological development process are improved, and the forward-looking and reliability of cell pathological change analysis are enhanced.

[0065] Embodiment III Please refer to Figure 3 , Figure 3 which is another cell image analysis system based on geometric mean and graph structure disclosed in the embodiments of the present invention. Figure 3 The described cell image analysis system based on geometric mean and graph structure is applied in a data processing system / data processing device / data processing server (wherein, the server includes a local processing server or a cloud processing server). As Figure 3As shown, the cell image analysis system based on geometric mean and graph structure may include: A memory 301 storing executable program codes; A processor 302 coupled to the memory 301; Wherein, the processor 302 calls the executable program codes stored in the memory 301 to execute the steps of the cell image analysis method based on geometric mean and graph structure described in Embodiment 1.

[0066] Embodiment 4 An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program causes a computer to execute the steps of the cell image analysis method based on geometric mean and graph structure described in Embodiment 1.

[0067] Embodiment 5 An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute the steps of the cell image analysis method based on geometric mean and graph structure described in Embodiment 1.

[0068] The above describes specific embodiments of this specification, and other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve the desired results. Additionally, the processes depicted in the figures do not necessarily have to be performed in the particular order shown or in a sequential order to achieve the desired results. In certain implementations, multitasking and parallel processing are also possible or may be advantageous.

[0069] The systems, devices, modules, or units illustrated in the above embodiments may be specifically implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0070] For convenience of description, when describing the above devices, they are described separately as various units according to functions. Of course, when implementing this specification, the functions of each unit may be implemented in one or more software and / or hardware.

[0071] Those skilled in the art should understand that the embodiments of this specification can be provided as a method, a system, or a computer program product. Therefore, the embodiments of this specification can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0072] This specification is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of this specification. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks.

[0073] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks.

[0074] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are performed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks.

[0075] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0076] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0077] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0078] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0079] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0080] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0081] Finally, it should be noted that what is disclosed by a cell image analysis method and system based on geometric mean and graph structure disclosed in the embodiments of the present invention is only the preferred embodiments of the present invention, and is only used to illustrate the technical solutions of the present invention, rather than limiting it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for cell image analysis based on geometric mean and graph structure, characterized in that The method includes: Inputting a multi-channel fluorescence image into a cell recognition algorithm model to obtain a cell distribution map; Calculating a cell deviation parameter in the cell distribution map according to the geometric mean algorithm; Extracting cell map features corresponding to the cell distribution map and the cell deviation parameter based on a graph neural network; Predicting the cell pathological classification and change trend at a future time point based on the time series corresponding to the cell map features at multiple historical time points and based on an LSTM neural network.

2. The method for analyzing cell images based on geometric mean and graph structure according to claim 1, wherein The inputting a multi-channel fluorescence image into a cell recognition algorithm model to obtain a cell distribution map includes: Inputting the cell fluorescence image of each channel into a trained cell image segmentation algorithm model to obtain a cell distribution map in the cell fluorescence image of each channel.

3. The cell image analysis method based on geometric mean and graph structure according to claim 2, wherein The cell distribution map includes a plurality of cell images and the cell type corresponding to each cell image.

4. The method for analyzing cell images based on geometric mean and graph structure according to claim 3, wherein, The calculating a cell deviation parameter in the cell distribution map according to the geometric mean algorithm includes: For each cell image in the cell distribution map, calculating the geometric mean of the image deviation parameters of the cell image in the cell fluorescence images of all channels to obtain the image average parameter corresponding to the cell image; Calculating the geometric mean of the position distances between the image position of the cell image and the positions of a plurality of reference cell positions to obtain the position average parameter corresponding to the cell image; Calculating the weighted sum average of the image average parameter and the position average parameter to obtain the cell deviation parameter corresponding to the cell image.

5. The method for analyzing cell images based on geometric mean and graph structure according to claim 4, wherein The image deviation parameter is the difference between the image parameter of the cell image and the standard parameter value; the image parameter includes at least one of image brightness, image color value, image morphology, image texture, and image spatial distribution parameter.

6. The method for analyzing cell images based on geometric mean and graph structure according to claim 4, wherein The reference cell position is the position of the cell image of the nearest preset dangerous cell type, the position of the cell image of the nearest same cell type, or the position of the cell image at the center of a preset cell region.

7. The method for analyzing cell images based on geometric mean and graph structure according to claim 1, wherein The extracting cell map features corresponding to the cell distribution map and the cell deviation parameter based on a graph neural network includes: Taking each cell image in the cell distribution map as a graph node, and taking the cell deviation parameter corresponding to the cell image corresponding to each graph node as the graph node feature parameter; For any two of the graph nodes, calculating the ratio between the cell deviation parameters respectively corresponding to the two graph nodes to obtain the node adjacency feature between the two graph nodes; Determining all the graph nodes and the corresponding node adjacency features between them as the graph network corresponding to the cell distribution map; Inputting the graph network into a trained graph neural network to obtain the cell map features corresponding to the graph network; the graph neural network is trained by a training data set including a plurality of training cell graph networks and corresponding feature annotations.

8. The method for analyzing cell images based on geometric mean and graph structure according to claim 1, characterized in that The predicting the cell pathological classification and change trend at a future time point based on the time series corresponding to the cell map features at multiple historical time points and based on an LSTM neural network includes: Sort the cell graph features of the same patient at multiple historical time points from the earliest to the latest time point to obtain a feature time series sequence; Input the feature time series sequence into a trained LSTM neural network to obtain multiple cell pathological types corresponding to the output future time points; Determine the pathological change trend corresponding to the patient according to the type changes of the multiple cell pathological types corresponding to the future time points.

9. A cell image analysis system based on geometric mean and graph structure, characterized in that, The system includes: An identification module for inputting a multi-channel fluorescence image into a cell recognition algorithm model to obtain a cell distribution map; A calculation module for calculating the cell deviation parameter in the cell distribution map according to the geometric mean algorithm; An extraction module for extracting cell graph features corresponding to the cell distribution map and the cell deviation parameter based on a graph neural network; A prediction module for predicting the cell pathological classification and change trend at future time points based on an LSTM neural network according to the time series sequence corresponding to the cell graph features at multiple historical time points.

10. A cell image analysis system based on geometric mean and graph structure, characterized in that, The system includes: A memory storing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory and executes the cell image analysis method based on geometric mean and graph structure according to any one of claims 1-8.

Images